Optimisation of an enzymatic method to obtain modified artichoke pectin and pectic oligosaccharides using artificial neural network tools. In silico and in vitro assessment of the antioxidant activity

Abstract

An enzymatic procedure to obtain modified artichoke pectin and pectic oligosaccharides (POS) (Mw 100–0.3 kDa) has been optimised through an experimental design analysed by artificial neural networks (ANN; R2 0.99), leading to high yields of these products (65.9 ± 2.1 mg 100 mg−1 pectin) at optimal conditions (pH 4.41, reaction time 0.9 h, enzyme dose 17.1 U g−1 pectin), reaching a maximum theoretical desirability of 0.98. Desirability function, variable importance and sensitivity analysis were performed to interpret ANN while residual analysis demonstrated its high predictive power. Hydrolysates were purified by ultrafiltration and retentate and permeate fractions were characterised by MALDI-TOF-MS. Oligosaccharides from di- to hexasaccharides corresponding to galacturonic acid (GalA) oligomers that may be attached to neutral sugars and ferulic acid were determined, and their potential free radical scavenger activity was calculated using an in silico model (72–98% probability). The presence of specific structures in permeate (high free GalA content, GalA oligomers attached to xylose, ferulic acid or rhamnose and arabinose) and retentate fractions explained differences observed in their in vitro antioxidant activities (135.6 and 32.1 μmol Trolox g−1, respectively). The combination of in silico and in vitro methods allows establishing structure-activity relationships for modified pectin and POS fractions.